Dear FSL experts,
As advised previously in the mailing list, prior to performing group ICA on my dataset,
I ran a denoising scheme, a variation of CompCor, to remove physiological noise.
After running group ICA both ways, with/without the denoising scheme, I realized that ICA components without denoising scheme actually
produced more favorable components - Comparing against standard templates for RSNs, denoised group ICA couldn't seem to produce as good-fit components.
This got me perplexed, as I expected that denoising prior to group ICA would actually help in finding clearer sources.
Does denoising distort the raw signal in some way and result in poorer results?
Thanks for the insight!
|